yoav_hollander
The V&V method – A step towards safer AGI
About “The coming AI hackers” paper
Estimating the residual risk of ADAS/AV
GPT-3 and Verification
Autonomy markets and their potential bugs
https://blog.foretellix.com/2017/07/06/where-machine-learnin...
M-SDL, the autonomous vehicles verification language, is now open
Why testing the typical is not enough for Autonomous Vehicles
Don't Stay in Monte Carlo (for Autonomous Vehicles Verification)
Moving from autonomous vehicles verification to regulation
Bridging autonomous vehicles verification and regulation
Running autonomous vehicle scenarios: Some problems
What to do about the next 500 fatal autonomous vehicle crashes
What to do about the next 500 serious AV accidents
The Uber accident and the bigger picture
How to write autonomous vehicle scenarios
Verifying how autonomous vehicles behave *during* accidents
Verifying how autonomous vehicles behave *during* accidents
On Mobileye's formal model of autonomous vehicle safety
Using program induction for verification – will this work?
Autonomous vehicle verification updates: Waymo, Apple and more
Autonomous vehicle verification updates: Waymo, Apple and more
What's new in autonomous vehicle verification – part two
What's new in autonomous vehicle verification – part two
[1] http://blog.foretellix.com/2017/07/06/where-machine-learning...
The only issue is that there is no easy, _natural_ way to do it. For instance, consider the various attempts at adding safety rules to an RL ANN (depicted in fig. 2 in the paper). Say that (in the context of an ANN controlling an Autonomous Vehicle) your ANN decided to do something on the freeway, but the safety rules say "no". There is no easy way to gracefully integrate the rule and ANN: One way is for the rule to disable the ANN's output at this point, take full control and decide what the AV _should_ do. But this leads to duplication and complexity.
So the four solutions I describe take various ways to avoid this problem. They all "work" in a sense, but none does real "integration" of the ANN and the rules (the shield synthesis solution perhaps comes closest). And it looks like you have to invent this kind of solution anew for every new instance of connecting-ANN-to-rules.
And this was just "inserting rules during execution". Then there is the issue of "verifying via rules", and "explaining the rules". It is tough, and I am wondering if there could be some conceptual breakthrough which would make it somewhat easier.